When Astrobotic’s Griffin lander descends to the lunar surface, it will precisely target a small landing ellipse (a small area where it might land) and autonomously maneuver to avoid hazards such as rocks bigger than 25cm and slopes greater than 15°. In last month’s blog post, we introduced the landing sensor package and the concept of map registration – a technique that matches (“registers”) a location in an in-flight image to the same location on a map.
This week, an Astrobotic team led by Kevin Peterson is headed out to Masten Space Systems, located at the Mojave Air and Space Port in Mojave, CA, to fly the landing sensor package and software system on the Masten Xombie suborbital rocket. The first flight will operate the system in an open-loop mode, where Astrobotic’s sensor package captures the same data it would use for an autonomous landing, but without actually controlling the vehicle. The second and third flights, slated for later this spring, will be closed-loop flights where Astrobotic’s landing software uses the sensor-package data in real time to guide the vehicle’s landing.
Terrestrial simulation of the landing task requires creativity. Over the last year, Astrobotic has used a variety of test environments to exercise the landing system components and gather data about their operation.
On a Zipline
The idea behind the zipline tests is to roughly approximate the trajectory of the final lunar descent in an open-loop test. We started the zipline trajectory 50m above the ground and then moved the sensor package steadily along the line’s downward arc. We tracked the sensor package from the ground with visual surveying tools to provide a known-accurate description of the trajectory for testing the visual odometry system.
On a Large RC Helicopter at Virginia Tech
Flying on a large remote-control helicopter subjects the sensor suite to a more realistic environment, including vibration, to ensure operation of all sensors and electronics. We controlled the trajectory of the helicopter to simulate landing on the surface of the Moon. RC helicopters are great platforms for prototyping closed-loop descent since, unlike most full-sized helicopters, they can be controlled by the landing software and, unlike RC planes, they can follow a path similar to a propulsive landing. We will fly the landing system closed-loop on this RC helicopter before landing the Masten vehicle.
On a Plane
The main limitation of the RC helicopter is that it can’t fly high enough to simulate the challenges the landing system will face early in the lunar descent. To collect data for open-loop testing of high-altitude map registration, we flew a camera comparable to our sensor package on a small airplane. We collected high-altitude photographs of the ground with accurate GPS locations for testing the map registration software’s ability to accurately determine the camera’s location by comparison to Earth satellite imagery.
On a Full-Scale Helicopter
The full-scale helicopter test brought our open-loop testing one step closer in realism to a lunar landing. While the zipline and RC flights were limited to about 60m above the ground, the helicopter can fly as high as 3,000m. We flew trajectories from 1,500m altitude down to 30m to simulate the lander’s descent from high altitude where stereo doesn’t work to low altitude where stereo becomes important. We used GPS as the known-accurate altitude and location to test our navigation software’s ability to accurately track its own altitude and location using data from our sensor package in a simulated lunar descent.
The open-loop Masten flights have three purposes. They will verify that all sensors function properly in the rocket flight environment, including shake, vibration, and high speeds; confirm that the Astrobotic and Masten coordinate systems are the same; and provide data to evaluate performance of map registration, stereo mapping, and hazard detection prior to closed-loop flights.
The closed-loop flight, slated for late spring, will confirm that Griffin can register to terrain (with altitude and location estimates within 10m at high altitude and 2m at low altitudes), detect hazards such as rocks bigger than 25cm and slopes greater than 15°, compute a safe landing location from hazard data, and reliably control the vehicle, all during real-time flight.
The upcoming flight campaign will include testing, tethered flights, open-loop flight, and post-flight analysis, over the course of several weeks in early to mid-February. Watch the Astrobotic Facebook and Google+ feeds for regular updates as the flight campaign progresses.